kendreaditya/lcsearch

lcsearch — LeetCode search & read

Read-only CLI for leetcode.com. Wraps scripts/lcsearch.py, a single-file Python tool that hits LeetCode's REST problem catalog and a small whitelist of public GraphQL operations. No login, no cookies, no API key.

Invocation

The script is on $PATH as lcsearch (symlinked from ~/.config/scripts/). Use that short form in examples. Falls back to python3 scripts/lcsearch.py from a fresh environment. Standard library only — no pip install step.

lcsearch --help

The first search or list call populates a 24h disk cache at ~/.cache/leetcode/problems.json (~3MB) — the full problem catalog from /api/problems/all/. Subsequent searches are instant.

Commands

daily — today's daily coding challenge

lcsearch daily [--json]

Returns the problem id, title, difficulty, and URL for today's daily.

problem — full problem statement

lcsearch problem <slug-or-id-or-url> [--json]

Accepts any of:

• Title slug: two-sum
• Frontend problem number: 1
• Full URL: https://leetcode.com/problems/two-sum/

Renders the full HTML problem statement (converted to plain text/markdown), tags, like/dislike counts, AC rate, hints, example test cases, available code-snippet languages, and similar-question links.

lcsearch problem two-sum
lcsearch problem 56
lcsearch problem https://leetcode.com/problems/merge-intervals/

search — search the problem catalog

lcsearch search "<query>" [--difficulty Easy|Medium|Hard] [--paid|--free] [--limit N] [--json]

Substring-matches on title and slug across the full catalog (~3,900 problems). Order is the LeetCode REST default (newest first). Also accepts a bare frontend id as the query.

lcsearch search "two sum"
lcsearch search "binary tree" --difficulty Hard --free
lcsearch search 56

Note: server-side filtered search (problemsetQuestionListV2) is auth-walled by LeetCode. This skill matches client-side over the public REST catalog, which is enough for most lookups but doesn't support filtering by tag.

list — browse the catalog with filters

lcsearch list [--difficulty Easy|Medium|Hard] [--paid|--free] [--limit N] [--json]

Same shape as search but without a keyword. Useful for skimming newest-added problems.

discuss — search the Discuss forum

lcsearch discuss "<query>" [--tag SLUG ...] [--order hot|recent|votes] [--limit N] [--json]

Searches the LeetCode Discuss forum (ugcArticleDiscussionArticles) by keyword. Each post lists title, author, topicId, ♥ upvotes, 💬 top- level comment count, 👁 hit count, tags (with * marking COMPANY tags), the canonical URL, and a wrapped summary.

lcsearch discuss "waymo interview"
lcsearch discuss "system design" --tag google
lcsearch discuss "compensation" --order recent --limit 10

--order accepts hot (default), recent, or votes.

feed — browse the Discuss feed by tag

lcsearch feed [tag-slug] [--order hot|recent|votes] [--limit N] [--json]

Like discuss with no keyword. Without [tag-slug] returns the global hot/recent/votes feed (3000 posts total per LeetCode). Pass a tag slug to filter — e.g. google, amazon, interview, career, compensation. Find available tags via lcsearch tags.

lcsearch feed google --order recent --limit 5
lcsearch feed compensation --order hot
lcsearch feed                              # global feed

read — full discuss post + comments

lcsearch read <topic-id-or-url> [--order best|hot|recent|votes] [--limit N] [--json]

Accepts a bare numeric topicId or a /discuss/topic/<id> URL. Renders the title, metadata, full body, and the top --limit comments. Post-slug URLs (/discuss/post/<slug>/) do NOT carry a numeric id — find the topicId via lcsearch discuss "<title fragment>" first.

lcsearch read 7949696
lcsearch read https://leetcode.com/discuss/topic/7949696/

Comment ordering vocabulary differs from the feed (best is the LeetCode default; recent maps to newest_to_oldest, votes maps to most_votes). The --order flag handles the translation.

comments — just the comments

lcsearch comments <topic-id-or-url> [--order ...] [--limit N] [--json]

Same as read but skips the post body — useful for re-checking replies on a thread you've already read.

company — company list metadata

lcsearch company <slug> [--json]

Returns metadata for a /company/<slug>/ page: name, total question count, collect count, the sub-list slugs (<slug>-thirty-days, <slug>-three-months, etc.), curated roles, and a clear note that the actual problem list is paywalled by LeetCode Premium.

lcsearch company waymo
lcsearch company google
lcsearch company meta

The actual question list for a company (favoriteQuestionList) is premium-only — anonymous calls return an empty list. This skill deliberately surfaces only the public metadata.

tags — hot discussion tags

lcsearch tags [--json]

Lists currently-trending Discuss tags with sample posts. Use this to find a valid --tag value for feed.

export — dump a discuss post to md or json

lcsearch export <topic-id-or-url> [--format md|json] [--limit N]

Writes a self-contained Markdown (default) or JSON document to stdout. --limit caps comment count. Useful for archiving interview-experience threads you want to keep around.

lcsearch export 7949696 --format md > waymo-interview.md
lcsearch export 7949696 --format json > waymo-interview.json

export-problem — dump a problem to md or json

lcsearch export-problem <slug-or-id> [--format md|json]

Writes the full problem (statement, hints, similar questions, all code snippets) to stdout.

lcsearch export-problem two-sum > two-sum.md
lcsearch export-problem 1 --format json > two-sum.json

Common flags (work on every command)

• --json — emit structured JSON instead of pretty text. Pipe into jq or load into Python.
• --limit N — cap output to N items (problems, posts, or comments).
• -v / --verbose — log fetched URLs to stderr for debugging.

Common flags must appear after the subcommand name, e.g. lcsearch discuss "google" --limit 5.

How to choose the right command

| User intent | Command | |---|---| | "What's today's LeetCode daily?" | daily | | "Show me LeetCode {problem name/number}" | problem {name} | | "Find LeetCode problems about {topic}" | search "{topic}" | | "What are Hard problems on tree-DP?" | search "tree" --difficulty Hard (then filter mentally — tag filter is auth-walled) | | "What's been posted on LeetCode discuss about {company}?" | discuss "{company}" or feed {company-slug} | | "Show me the {company} discuss feed" | feed {slug} (slug from tags) | | "Read the LeetCode post at {url}" | read {url} | | "What's hot on LeetCode discuss right now?" | feed --order hot | | "Recent interview experiences at Google" | feed google --order recent | | "Save this discuss thread" | export {topic-id} --format md > file.md | | "Save this problem with starter code" | export-problem {slug} --format md > file.md | | "What companies has LeetCode curated lists for?" | company {slug} (try common slugs: google, meta, amazon, microsoft, apple, nvidia, waymo, openai, anthropic) |

When the user gives a leetcode.com/problems/... URL → problem. When they give a leetcode.com/discuss/topic/<id> URL → read. When they describe a topic → search (problems) or discuss (forum). When they name a company → feed <slug> (forum) or company <slug> (curated list).

Cross-skill pairings

• interviewcoder — structured leetcode-style question writeups sourced from 1point3acres, tagged by company/stage/period. Use it for the question text, then lcsearch discuss "{company}" for candidate-side commentary.
• blind — anonymous Blind chatter on the same companies. Layoffs, severance, comp data points. Pair both when a user is prepping for a specific employer.
• forage / reddit — broader experiential search if neither of the above has hits.

Data schema (JSON output)

Problem (from search, list, daily): id, frontend_id, title, title_slug, difficulty, paid_only, ac_rate, total_accepted, total_submitted, url, tags[].

ProblemDetail (from problem, export-problem): wraps Problem plus content_html, content_text, hints[], likes, dislikes, similar_questions[], example_testcases, code_snippets[] (each with lang, lang_slug, code).

DiscussPost (from discuss, feed): topic_id, uuid, title, slug, summary, author, author_slug, created_at (ISO8601), hit_count, upvotes, comment_count, tags[] (each {name, slug, type} — type is COMPANY for company tags, else null), url.

Article (from read, export): post (DiscussPost without inlined body), content (full body), comments[], total_comment_count. Each Comment has id, author, content, created_at, vote_count, num_children, pinned.

CompanyInfo (from company): slug, name, description, question_count, collect_count, is_public, favorite_type, last_modified, creator, generated_slugs[], position_role_tags[], url.

Limitations

• No login. LeetCode auth-walls a lot — problemsetQuestionListV2 (filtered problem search by tag), favoriteQuestionList for company slugs (the actual problems on a company page), submissions, your own progress, and anything personalized. This skill deliberately exposes only what works anonymously.
• Problem-list filtering is client-side over the cached REST catalog. Difficulty and paid/free filters work; tag filtering does not (the tag list isn't in the REST payload).
• First page only for discuss and feed per call (50 max per LeetCode's first-page cap). Pagination beyond skip=0 is not exposed.
• No write operations. No posting, no commenting, no voting, no bookmarking, no submissions.
• No replies. read returns top-level comments only. The reply count (num_children) is shown, but to read the replies the user has to open the URL.
• No reactions detail. Only upvote count is surfaced; LeetCode's full reaction palette (downvotes etc.) is in the raw payload but filtered out.

Error handling

The script exits with:

• 0 on success
• 1 on errors (network, GraphQL error, parse failure)
• 2 on "not found" — bad problem slug, bad topic id, empty search

GraphQL operations that turn out to require auth will fail with a clear error: requires authentication message naming the operation.